MATH - Equating Mirrorless AF Points to DSLR AF Points

[bmwzimmer]

I posted this on DPreview. I have a feeling I'm going to get grilled for it but it is pretty damn accurate in my experience....

The General formula to determine how well a Mirrorless Camera Autofocuses compared to a DSLR...

1.) Add the Number of Phase Detection AF points with Contrast Detection Points
2.) Take the Square Root of that Number
3.) Spread it evenly across the Frame
4.) Pretend this is the number of AF point on a DSLR (that are NOT CROSS TYPE on a Modern DSLR (Nikon D7200 or D750 which have more modern better non-cross type points than Canon's 5D/6D))

For the Case of the A7ii, it's 117 + 25 = 142
The Square root of that is 11.9 or about 12 points
Thus, the A7ii will perform like a modern DSLR with 12 AF points that are not cross type.

For the Case of the A6000, the sq root of (179 + 25) = 14 non cross type points but spread closer together since it's a smaller sensor so the AF is somewhat decent.

For the A7Rii, you have the sq root of (399 + 25) = 20.5 non cross type points across the frame so performance will be similar to an A6000.

______________________________________________

Here's one more Method that works as well

1.) Take the Square Root of that Number of Phase Detection AF points and the square root of Contrast Detection Points
2.) Spread it evenly across the Frame
3.) The sq root of Phase Detection points is equivalent to Non Cross type AF points on a Modern DSLR.
4.) The sq root of Contrast Detection AF points equals the number of Cross Type AF points on a 5-6 year old DSLR (i.e. Canon 50D)

For the Case of the A7ii, it's 117 & 25. So you get 11 AF points (modern DSLR non cross type) and 5 cross type AF points (5-6 year old DSLR Cross Type) for a total of 16 AF points

For the Case of the A6000 (179 & 25), you get 13 AF points, 5 Cross type across a smaller sensor for a total of 18 AF points so AF tracking is somewhat acceptable

For the A7Rii, (399 & 25), you get a respectable 20 AF point and 5 Cross type for a total of 25 AF points across a larger sensor so AF should be as good as the A6000.

For the NX1, (205 & 209), you get 28pts, 14 of which are cross type


In All seriousness, this has been my experience so far after spending time with the A6000, NX1, NX500, A7, and X-T1 and I really think it's a pretty accurate formula.

 

[meywd]

can you explain the math, like why you take the square root?

 

[3kramd5]

can you explain the math, like why you take the square root?

The tried and true method of making sh*t up.

It's not a rigorous model, it's a correlation to a subjective measure.

 

[bmwzimmer]

can you explain the math, like why you take the square root?

The tried and true method of making sh*t up.

It's not a rigorous model, it's a correlation to a subjective measure.

Yup, like many Engineering and materials equations, coefficients are created based off tests and observations. This is just my observation for what i've seen to date and correlating it to the new A7Rii.

 

[3kramd5]

can you explain the math, like why you take the square root?

The tried and true method of making sh*t up.

It's not a rigorous model, it's a correlation to a subjective measure.

Yup, like many Engineering and materials equations, coefficients are created based off tests and observations. This is just my observation for what i've seen to date and correlating it to the new A7Rii.

I have no issues with observing and fitting curves to those observations. For sure rules of thumb are used extensively in engineering, even in the analysis side of things wherein margin (one of countless examples) may be based in part on rules-of-thumb about how proximity of a hole to the edge of a part affects stress.

I do question how to correlate to a subjective measure such as AF performance, though. Not saying it can't be done, but it's certain difficult without actively using a multitude of cameras in a controlled environment.

 

[Hillsilly]

Just wanted to point out that:-

1. DSLR AF points are heavily clustered around the centre, whereas mirrorless AF points cover most of the sensor;
2. Mirrorless AF sensors are more accurate (even moreso if you have to focus and recompose your DSLR due to the problems associated with much smaller AF coverage.)

I'm not very good at maths, but I get the impression that this formula suggests that mirrorless AF is inferior to DLSR AF? I would have thought it would be the other way around, with mirrorless AF far superior in every aspect except AF tracking.

 

[bdunbar79]

Just wanted to point out that:-

1. DSLR AF points are heavily clustered around the centre, whereas mirrorless AF points cover most of the sensor;
2. Mirrorless AF sensors are more accurate (even moreso if you have to focus and recompose your DSLR due to the problems associated with much smaller AF coverage.)

I'm not very good at maths, but I get the impression that this formula suggests that mirrorless AF is inferior to DLSR AF? I would have thought it would be the other way around, with mirrorless AF far superior in every aspect except AF tracking.

Why do you say mirrorless sensors are more accurate? Traditionally they have been mostly contrast-detect which while more accurate than phase, is much slower.

Secondly, why do you think phase-detect sensors on sensor are more accurate? Traditionally they have been less accurate. Then you have the whole precision thing, with cross-types and dual cross-types, where up until now, mirrorless cannot compete. Think 1Dx vs. A7R II. Which would you pick if you had one NFL game where you were going to be paid $1 million.

Tracking is horrendous in mirrorless vs. DSLR's.

 

[bmwzimmer]

It takes longer to focus and its not nearly as snappy. It may be more accurate on paper but what good is it when it hunts more or just unable to lock focus. Transitions from near to far focus is much slower and it AF performance suffers when objects move towards or away from the camera in strong backlit conditions (side to side is usually fine). All this is good lighting. In poor lighting, its much much worse.

 

[Valvebounce]

Hi meywd.
Don't you know that every formula worth its salt has to have as square root in it somewhere! : ;D
On a more serious note, do the focusing pixels leave a hole in the image or do they record light colour level too? I'm guessing if they don't it is no worse than a dead pixel, except that there are many more focusing pixels than you would want dead pixels!

Cheers, Graham.

can you explain the math, like why you take the square root?

 

[RLPhoto]

Just wanted to point out that:-

1. DSLR AF points are heavily clustered around the centre, whereas mirrorless AF points cover most of the sensor;
2. Mirrorless AF sensors are more accurate (even moreso if you have to focus and recompose your DSLR due to the problems associated with much smaller AF coverage.)

I'm not very good at maths, but I get the impression that this formula suggests that mirrorless AF is inferior to DLSR AF? I would have thought it would be the other way around, with mirrorless AF far superior in every aspect except AF tracking.

Why do you say mirrorless sensors are more accurate? Traditionally they have been mostly contrast-detect which while more accurate than phase, is much slower.

Secondly, why do you think phase-detect sensors on sensor are more accurate? Traditionally they have been less accurate. Then you have the whole precision thing, with cross-types and dual cross-types, where up until now, mirrorless cannot compete. Think 1Dx vs. A7R II. Which would you pick if you had one NFL game where you were going to be paid $1 million.

Tracking is horrendous in mirrorless vs. DSLR's.

The only on sensor phase detect AF I've been impressed with was the 70D. It really is sweet but even then it's slow.